How do diverse infections reprogram the ileal epithelium and epithelial-immune interactions?

An atlas built from six infections and naive mice

GutPath was assembled to answer a gap the authors describe directly: single-cell studies of the gut have concentrated on the large intestine, and infection-induced immune processes in the ileum are underrepresented. The team therefore sampled the ileum, its draining mesenteric lymph node (MLN) and, for one pathogen, the duodenum, using CITE-seq, which reads messenger RNA and about 100 surface proteins from the same cell.

The panel spans distinct phyla: segmented filamentous bacteria (a gram-positive commensal), Yersinia pseudotuberculosis (gram-negative bacterium), Nippostrongylus brasiliensis (helminth), Candida albicans (fungus), Cryptosporidium parvum (apicomplexan parasite) and murine norovirus strain CR6. All were given orally except Nippostrongylus, which was injected subcutaneously as third-stage larvae, and tissue was collected during the first week of infection.

After quality control, 505,956 cells were retained: 184,973 from the MLN and 320,983 from the ileum. Replicate mice per condition ranged from two to three, so the atlas is broad in conditions but shallow in biological replication, a point that matters when reading any single condition-specific difference.

Ninety-one ileal cell states, annotated in layers

The atlas defines 91 transcriptional cell states in the ileum and 49 in the MLN, organized from broad lineages down to refined states. In the ileum, epithelial cells alone account for 148,453 cells, stromal for 18,570, myeloid for 16,542, B cells for 53,934, innate lymphoid cells for 21,946, CD8 and gamma-delta T cells for 38,594, and CD4 and NKT cells for 15,228.

This resolution is the paper's practical contribution. The authors generated marker profiles at coarse, intermediate and refined annotation levels, plus condensed random-forest marker sets, so that other datasets can be labeled by transfer rather than annotated from scratch. The claim is about usability: the atlas is offered as a reference for mucosal single-cell work, not only as a description of these six infections.

One structural caveat belongs here. The authors note that their data are biased toward epithelial cells, which may have increased statistical power for detecting differentially expressed genes in that population. Epithelial findings therefore rest on more cells than findings in fragile populations such as eosinophils and neutrophils, which the authors say may be underrepresented.

The atlas recovers expected immune hallmarks

Before treating novel findings as credible, the authors checked whether the atlas reproduces established infection biology. Using differential abundance testing of cell neighborhoods defined by nearest neighbors, they found tuft and goblet cell neighborhoods enriched in Nippostrongylus infection but not in the other infections, matching the tuft and goblet cell hyperplasia expected in helminth infection.

Recruitment of monocytes, neutrophils and other myeloid effectors appeared as a dominant feature of many infections, and Yersinia specifically expanded neutrophils in the lamina propria while Nippostrongylus expanded tuft cells and innate lymphoid cells. Because these patterns align with prior expectations, they function as an internal validation step: the pathogen-specific signals described next are more plausibly biology than processing artifact.

Infections disturb solute transporters along enterocyte maturity

Because sample preparation captured the full enterocyte lifespan, from crypt stem cells through transit-amplifying cells to mature absorptive cells at the villus tip, the authors could order cells along pseudotime and ask where transport genes go wrong. They tracked 82 solute transporter genes and found a structured expression pattern from crypt to villus tip, with 29 of the 82 dysregulated during one or more infections.

This is a distinct result from the immune cell findings: it concerns the absorptive machinery of the epithelium rather than immune infiltration. The authors frame the transporter dysregulation as a potential molecular mechanism for pathogen-driven malabsorption, drawing a link to the malabsorption seen in Crohn's disease and infection, but they present it as a hypothesis that merits further investigation rather than a demonstrated cause.

Helminth infection reshapes enterocyte metabolism

Among the infection conditions, enterocytes from Nippostrongylus-infected mice showed the most pronounced shifts in predicted metabolic tasks, particularly lipid pathways. These included reduced synthesis of secondary bile acids and altered metabolism of fatty acids such as elaidate, linoleate, palmitate and arachidonate, with fatty acid processing genes increased relative to all other conditions.

The authors localized this effect: the median difference in fatty acid-related activity scores between infected and naive samples was highest in enterocytes, regardless of maturity, and in goblet cells. Cell-cell communication analysis predicted that infected enterocytes receive interleukin-4 and interleukin-13 signals, consistent with a TH2-dominant response, and produce ligands including FGF15 and CX3CL1 that were inferred to act on other populations.

Two checks support the interpretation. Genes induced in enterocytes during infection, including Pla2g4c, were also upregulated in enteric organoids stimulated with IL-4, and the authors note an association between worm burden and fatty acid levels. That association is correlative, and because Nippostrongylus establishes infection in the proximal small intestine, the metabolic change seen in the distal ileum is evidence that effects extend beyond the site of infection.

A Yersinia-linked enterocyte state tracks pathology

The discovery the authors highlight most is a distinct enterocyte transcriptional state found during Yersinia pseudotuberculosis infection, labeled Yps enterocytes, alongside a Stat1+ enterocyte state and an inflammatory trophocyte population expressing Il6, Ccl2 and Cxcl1. Using spatially resolved single-cell transcriptomics, they show that some enterocyte states colocalize precisely with tissue pathology, while others are regional, sitting adjacent to pyogranulomas, or spread across the whole tissue.

The proposed mechanism is signaling-based and explicitly inferential: the data suggest the Yersinia-associated state may be driven by IL-22/STAT3 signaling to produce IL-18, which then signals through IL-18RAP on innate lymphoid cells, NK cells and NKT cells residing in the pyogranuloma. This is a predicted communication chain from expression data, not a perturbation experiment.

The spatial result also carries a broader observation. Although immune infiltrates and proinflammatory gene expression concentrated around pyogranulomas, genes such as Saa1, Stat3 and Slc40a1 were widely expressed in enterocytes across the entire distal small intestine. The authors call this a ripple effect, and they leave open whether focal cytokine induction is sufficient to drive tissue-wide downstream targets or whether enterocytes themselves propagate inflammatory signals over distance.

One early time point limits what the atlas can claim

Each infection was sampled at a single early time point chosen to coincide with peak infection. The authors state that differences in how infections establish and progress, and how adaptive components are recruited, cannot be fully explained without sampling along the time course, and that the persistence of the transcriptional programs and their effect on effector or memory T cell development remain unclear. Infectious doses were standard for the field, but dose effects were not tested.

A second boundary is sex. All mice were female C57BL/6J, so the authors cannot confirm that the conclusions apply to male mice. Together with the two-to-three replicate mice per condition, these constraints mean the atlas is strongest as a reference map and hypothesis generator, and weaker as a source of precise effect sizes for any single infection.

The atlas shifts attention to the epithelium

The most consequential move in this paper is not the cell count but the direction of attention. Immune responses to infection were already partly known, and GutPath reproduces them; what it adds is evidence that enterocytes carry pathogen-specific programs, that these programs vary by maturity along pseudotime, and that they can appear far from the site of pathogen load.

For readers deciding what to do with this, the atlas supports two concrete uses. First, its layered annotations and marker profiles allow label transfer into new mucosal single-cell datasets, which is where the resource saves the most work. Second, the genes spatially tied to Yersinia pathology, and the transporter and lipid-metabolism changes tied to infection, are the specific candidates the authors propose for follow-up as therapeutic targets or drivers of tissue remodeling. The priority is the epithelium and its signaling to immune cells, not another census of immune infiltration.

Sources used in this answer